AI Sprint

Services

Training makes the jobs you already do faster. A Sprint asks whether they’re the right jobs. We take one process that runs across a team, work out what’s actually wrong with it, and rebuild it with the people who run it.

Who it’s for

Teams where people are already handy with AI on their own jobs, and the business now wants to take one whole workflow and make it genuinely better. Work that passes between several people, not one person’s task.

Before the Sprint

Two things happen first.

1

Agree what this workflow is actually for. Speed, quality, volume, fewer mistakes. Pick what matters, because it decides what we optimise for and how we’ll know it worked.

2

Baseline it, starting as soon as we agree to go ahead rather than in week one. Whoever does the work keeps a rough note as they go, when they did it and roughly how long it took. Nobody can accurately remember this afterwards, and the baseline is what we measure the change against. Interviews can start early too, and the sooner they do the more real data we have.

Do this properly and the first week of the Sprint gets cut roughly in half.

The Sprint

You don’t get told the answer in week one. The first half is understanding how the work really runs. The second half is agreeing what it should be and building it.

A typical four weeks. Shown as the shape it usually takes, not fixed dates.

Time with us
Week 1
Interviews Two each. How the work really gets done, not how it is meant to.
Workflow map review One hour. Check we have captured how the work actually runs.
End of week 1 A map of how the work happens today

Week 2
Target state review One hour. What we think should change, and what the workflow looks like after.
Sign-off One hour, only if the first session needs changes.
End of week 2 An agreed target state, and what it takes to get there

Weeks 3 and 4
Week 3 Week 4
In your office Two to three days a week. Built, tested and refined by your team in your systems, then measured against your baseline.
Playback The new process, end to end, with the numbers.
End of week 4 The process running, the numbers, and a costed roadmap
What lands, and when
End of week 1 A map of how the work happens today

End of week 2 An agreed target state, and what it takes to get there

Week 3 Building and testing
End of week 4 The process running, the numbers, and a costed roadmap
  1. Week 1 About two hours each, plus a one-hour workshop. Virtual.

    Gather and map

    We interview everyone involved in the process, separately, twice. How the work actually gets done, not how the process document says it does, including the times it went wrong. Then one workshop to put what we heard in front of you and agree where the pain actually sits.

    You get

    A map of how the work happens, step by step. Every step with what goes in, what comes out, which system it happens in, how long it takes and how often.

    We need from you

    Two interviews each with everyone who touches the process, and the people who do the work in the workshop, not just whoever owns it.

  2. Week 2 One hour, or two if it needs a second look. Virtual.

    Agree the target state

    Every option that might address what we found, across people, process and technology. Not just the AI ones. Some of what comes out of this is a rule change or dropping a step entirely. We score them on impact, effort and cost, and put them to you alongside what we think the workflow should become. If you want it changed, we change it and come back.

    You get

    An agreed target state, with the map redrawn to show it, and the options behind it scored so you can see why.

    We need from you

    One hour to review it. A second only if you want changes made first. Otherwise this week is mostly us.

  3. Weeks 3 and 4 Two to three days a week, in your office.

    Build and measure

    We build the agreed changes on the real process and try them. Whatever shows measurable benefit against your baseline stays, whatever doesn’t gets dropped. Then the winner goes in place, running. The four weeks end with a playback: the new process run end to end in front of the people who will use it, with the numbers and the handover done in the room rather than by email.

    You get

    The process running differently and in place, evidence it worked against the baseline you set, and the method and templates so you can do the next one yourselves.

    We need from you

    More than any other phase. It’s being set up by your people, in your systems, on your real work. That’s the point. Doing it yourselves is how you can do the next one without us.

What you leave with

A process that runs differently, in place and working.

Evidence it worked, measured against the baseline you set at the start.

The method and the templates, so you can do the next one yourselves.

A costed roadmap of what to do next.

Afterwards

You get a recap of what was built, how it runs and what it saved, written down rather than left in someone’s head.

And I’m around for questions for a couple of weeks after the Sprint ends.

Format and price

A typical four weeks, sometimes less if your prep is done. £3,500 plus VAT.

The first two weeks are virtual. The last two are two to three days a week together in your office.

See what this looks like in practice

View all case studies

Hotel Mike

From two new retailers a month to six.

We redesigned and implemented the brand’s 13-step retailer-growth process across discovery, outreach, ordering and invoicing.

New retailers secured per month 2 → 6

Read the case study

Goodwill

Four days a month freed across five workflows.

A four-week Sprint redesigned how Goodwill finds prospects, analyses competitors, creates content and runs paid and organic marketing.

Personalised pitches prepared each week 20 → 60

Read the case study

Symbioss

Cold prospecting replaced by timely, relevant outreach.

A four-week Sprint built two connected workflows: monitoring that surfaces a real reason to make contact, and event prep that works on the day.

Watched continuously across her target market 11 signals

Read the case study

Thirty minutes, and you’ll know whether the problem is the tasks or the process.

Book a free AI review

An invented example, standing in for real client work. Every step as it actually runs, including the two that only happen half the time.

AttributeEnquiry arrivesWork out what they wantCheck availabilitySend a quoteChase it
What goes inEmail or web form, no fixed formatThe enquiry as written, plus whatever the last email saidJob dates and rough scopeScope, dates, day rateA sent quote with no reply
What comes outA new enquiry sitting in a shared inboxA short brief, when someone writes oneA yes, a no, or a maybe from two peopleA quote document, emailedA reply, or silence
Which systemOutlookNobody is sure. Notes, heads, sometimes TeamsShared calendar plus a resourcing spreadsheetWord, then OutlookOutlook, from memory
How long it takes2 minutes20 to 40 minutes10 minutes, longer if it needs the delivery team45 minutes5 minutes each time, three times on average
How oftenAround 40 a weekAround 40 a weekAround 30 a weekAround 18 a weekAround 12 a week
Fully happeningYesHalf. Only written down when the job looks bigYesYesHalf. Only when someone remembers it is theirs

The same map with two proposed steps added, then the options we scored against it. Figures are illustrative.

Attribute Enquiry arrivesWork out what they wantCheck availabilitySend a quoteChase it Triage and tag ProposedDraft the brief automatically Proposed
What goes inEmail or web form, no fixed formatThe enquiry as written, plus whatever the last email saidJob dates and rough scopeScope, dates, day rateA sent quote with no replyThe enquiry, unreadThe enquiry plus the tag
What comes outA new enquiry sitting in a shared inboxA short brief, when someone writes oneA yes, a no, or a maybe from two peopleA quote document, emailedA reply, or silenceEnquiry tagged by type, size and ownerA brief in the standard shape, ready to check
Which systemOutlookNobody is sure. Notes, heads, sometimes TeamsShared calendar plus a resourcing spreadsheetWord, then OutlookOutlook, from memoryOutlook, rules and an assistantAssistant, writing into the quote template
How long it takes2 minutes20 to 40 minutes10 minutes, longer if it needs the delivery team45 minutes5 minutes each time, three times on averageSeconds2 minutes to check, instead of 30 to write
How oftenAround 40 a weekAround 40 a weekAround 30 a weekAround 18 a weekAround 12 a weekEvery enquiryEvery enquiry
Fully happeningYesHalf. Only written down when the job looks bigYesYesHalf. Only when someone remembers it is theirsProposedProposed

The options, scored

The kinds of things that come up, across technology, process and people. Not a menu to choose from.

OptionWhere it sitsImpactEffortCost
Connect the quoting spreadsheet to the inbox, which has no ready-made connectorTechnology5 of 53 days29 pounds a month
Schedule the chase to run itself each morning rather than waiting for someone to start itTechnology4 of 51 day14 pounds a month
Drop the internal approval step on jobs under 2,000 poundsProcess4 of 5Half a dayNothing
Change the rule that every enquiry needs a call first, which nobody has questioned in yearsProcess3 of 51 dayNothing
Standardise the enquiry form so the quoting step gets what it needs first timeProcess4 of 52 daysNothing
Move availability checking to the delivery lead, who already knows the answerPeople3 of 5Half a dayNothing
Name one owner for the workflow after the Sprint, so it keeps being improvedPeople2 of 5Half a dayNothing
Book a free AI review